Data Modelling Best Practice for Survey

Hi,

I have a large dataset of over 1500 entries with 150 columns. It consists of personal particular questionaires, and dozen other questions composed of radio button (single input), checkboxes (multiple inputs), and textbox (open ended).

The questions can be broken down into 3 levels: Theme -> questions -> further details.
I.e. Customer experience -> Which newspaper do you read? -> when; where?

The output data is flat, the variables for radiobutton, and texbox questions are shared in a single column, while the checkbox questions created separate columns for its respective variables. See below.

My questions are:

1. Should I break down the flat database into its respective questions or by its theme?
2. How should I organize the nested structure?
3. Should I have the personal particulars as the master model?
4. Should I create a new column for all models to link it back to the master survey entry id? Where the new column is the same entry id?
5. Should I separate out the variables that share the same column? Some of my chart requires pivoting. I don't know what is the best way to do it.
6. Should I use binary format to manipulate the data or can I used the original variable names?

Thank you!